Wall function with k-w SST

A subforum specific to the development of the OpenFoam-based workbenches ( Cfd https://github.com/qingfengxia/Cfd and CfdOF https://github.com/jaheyns/CfdOF )

Moderator: oliveroxtoby

julieng
Posts: 40
Joined: Sun Nov 25, 2018 8:57 pm

Wall function with k-w SST

Postby julieng » Tue Dec 18, 2018 8:59 pm

Dear all,

Does CfdOF employs a wall function ? If yes which one?

https://www.cfd-online.com/Forums/openf ... a-sst.html

I understand that k-w SST without wall function is a ind of Low Re model needing a y+ <1 (or close to 1).
When used with a wall function (nutSpalartAllmarasWallFunction and omegaWallFunction) 30<y+<100-300.

What is your strategy concerning the mesh close to the wall ?

Best regards

Julien
m.cavallerin
Posts: 75
Joined: Wed May 30, 2018 6:59 pm

Re: Wall function with k-w SST

Postby m.cavallerin » Wed Dec 19, 2018 4:13 am

For what I know CFDOF allow the usage og the kOmegaSST model.

About mesh strategy at the wall?

Keep in mind the https://en.wikipedia.org/wiki/Law_of_the_wall

I'm not an expert (so I'm not academic) but that's what I understood about wallFunctions...

(BL: stands for boundary layer)
When you choose to model the BL so I mean you set an higher number of prism layer (let say 12to20 or more, depends on the problem) with the proper expansion factor to capture the gradients of your variables on the "y coordinate", you're not supposed to specify a wall function as you're setting up your simulation to resolve the BL: so it means that according to the Neumann problem, you have to specify the BC for all your variables near the walls: k, omega, nut, U, p, ecc... and the solver is able to compute properly the variables at each point of the mesh.
This strategy is intended if you want to CALCULATE PROPERLY the heat transfer coefficient, the drag forces, ecc because maybe you're designing a new stage for a SpaceX vector and you can't be so "interpolated"....

If you choose to not model the BL, (so you're using few prism layer that can have a y+ of 10-30) you are not so interested for the calculation of the variables near the wall...or better...
You can set up a mesh with few prism layer that model quite well the y+ law of the wall from 10 higher, but as you're using the wall functions formulation for the fluid at the wall: "the parabola" (of the Von Karman graph) is interpolated. So you're saying: "well, more or less it should be this...."
This strategy is intended if your problem is: "Well, I have an air flowrate that flows over a hot surface,what is the MEAN VALUE of my temperature at the outlets?"
understand that k-w SST without wall function is a ind of Low Re model needing a y+ <1 (or close to 1).
Is it true? are you able to model properly on a complex shape your mesh with a constant value of y+ equal to 1?
I think that as long as you're near the wall the interpolation (an error to the calculation) will be smaller.
I mean that if you start to interpolate from y+100 to the wall your error will be bigger compared to interpolate starting form a y+ of 10..and it should be smaller if you start from a y+ of 5 and smaller from 3... and so on...as along as you reach the wall, so why interpolate at this point?
julieng
Posts: 40
Joined: Sun Nov 25, 2018 8:57 pm

Re: Wall function with k-w SST

Postby julieng » Wed Dec 19, 2018 11:32 am

Thank you for your answer.

It is not clear for me if the actual version of cfdOF solver k-w SST use or not wall functions ? I think not but I need a confirmation.

To summarise, correct me if I am wrong (with the k-w SST turbulence model):
1- If I want to calulate fluxes, forces ... near the walls, I have to avoid wall functions and use instead BL with y+ close to 1.
2- If I want to estimate global variables on outlets or fluid velocities away from surfaces, I can use wall functions with a corase mesh and y+ close to 30 100 to save computation ressources.
- The solution 1 should be theorically always better whatever we want to calculate.

Best regards

Julien
User avatar
oliveroxtoby
Posts: 247
Joined: Fri Dec 23, 2016 9:43 am
Location: South Africa

Re: Wall function with k-w SST

Postby oliveroxtoby » Wed Dec 19, 2018 8:47 pm

julieng wrote:
Wed Dec 19, 2018 11:32 am
It is not clear for me if the actual version of cfdOF solver k-w SST use or not wall functions ? I think not but I need a confirmation.
Yes, the cases are created with wall functions (you can inspect the created 0/k, 0/omega and 0/nut for the details). This will give best results with 30 < y+ < 100.
m.cavallerin
Posts: 75
Joined: Wed May 30, 2018 6:59 pm

Re: Wall function with k-w SST

Postby m.cavallerin » Wed Dec 19, 2018 8:59 pm

It is not clear for me if the actual version of cfdOF solver k-w SST use or not wall functions ? I think not but I need a confirmation.
Now I have not the module installed but I remember so, in any case when you save your case to your folder you can edit the turbulencePrperties in your constant folder with:

Code: Select all

simulationType RAS;

RAS
{
    RASModel        kOmegaSST;

    turbulence      on;

    printCoeffs     on;
}
If I want to calculate fluxes, forces ... near the walls, I have to avoid wall functions and use instead BL with y+ close to 1
If I want to estimate global variables on outlets or fluid velocities away from surfaces, I can use wall functions with a corase mesh and y+ close to 30 100 to save computation ressources.
The solution 1 should be theorically always better whatever we want to calculate
The wall functions are "functions" that interpolate.

Good reading!!!

http://www.wolfdynamics.com/images/pdf/ ... ce_adv.pdf
Attachments
Screenshot from 2018-12-19 21-54-23.png
Screenshot from 2018-12-19 21-54-23.png (375.52 KiB) Viewed 253 times
Screenshot from 2018-12-19 21-53-29.png
Screenshot from 2018-12-19 21-53-29.png (348.08 KiB) Viewed 253 times